Ideas
Three questions run through everything I study. They look like three different topics. They're the same problem at three altitudes: a system designed for a world that no longer exists, and what it takes to redesign it.
The ideas
-
The Future of Work
The job was never the right unit of work. AI is proving it. When an agent does the execution, what exactly are we paying a person for? And pay is only the start of what has to be rebuilt.
-
The Human Side of AI
Your AI rollout isn't failing because the technology is hard. It's failing on trust. Your people don't know what it means for them, and uncertainty reads as threat. No one designed for that.
-
Engineering Performance
Your performance system was designed for people who don't exist. Here's how to build one for the people you actually have.
The Future of Work
A people problem that's actually a systems problem
What is a person worth when the agent does the work?
For a century, pay rested on a assumption: human effort produced organizational output, so you paid for the effort. Salary surveys, grade levels, job architectures, all of it was built for stable jobs where the human was the engine.
None of those conditions hold anymore.
Agents are absorbing execution. The humans who remain do something categorically different: they exercise judgment, they bear accountability, they design systems, they hold trust. The job is unbundling into agent-handled tasks and human-held responsibilities, and the old logic can't price the difference.
This is the part most leaders miss. Pay has always measured what the market charged. It has never measured what the work was worth. That gap was tolerable when the market moved slowly. In the agentic era it's a liability, because the market data you're benchmarking against is describing jobs that are dissolving in real time.
Pay is also the largest investment most organizations make in their people, and the least governed. We govern travel budgets more rigorously than we govern the decisions that set what millions of dollars of talent are paid. That was expensive before. Now it compounds.
The fix isn't a better survey. It's a different question. Stop asking what the market pays for this job. Start asking what this work is worth to us, and design the answer from the inside out.
Key ideas
The job was never the right unit, and now it's failing
Pay was always doing four jobs at once; grade level collapsed them into one number
Total Cost of Work: size human pay against the leverage a person orchestrates, not against peer grades
Pay for scarcity, not supply. For outcomes, not effort.
The Human Side of AI
A people problem that's actually a systems problem
The resistance to your AI rollout isn't resistance. It's data.
I studied 950 employees across the US and EU. Here's what the data actually shows. 44% show symptoms of what I call AI Angst, a specific mix of threat and uncertainty that predicts disengagement. 45% are hiding their AI use from the very companies that bought the licenses. And 70% of the barriers to adoption are clarity gaps, not capability gaps. People don't know if using AI is safe, sanctioned, or career-ending, so they either freeze or go underground.
None of that is a technology problem. It's a design problem. When you drop a powerful tool into a system that hasn't told people what it means for their job, their status, or their worth, fear is the rational response. The behavior you're calling resistance is people protecting themselves from a system that hasn't accounted for them.
The organizations getting this right aren't the ones with the best models. They're the ones designing the human side deliberately: clear rules, psychological safety to experiment, and honest answers to the question every employee is actually asking, which is "what does this mean for me?"
Manage AI from 30,000 feet and you get compliance without commitment. Design for the human underneath and you get adoption that sticks.
Key ideas
AI Angst: what it is, who has it, what it costs
Shadow AI: why your best people are hiding their most productive tool
Psychological Ergonomics: designing systems that work with human psychology, not against it
Human-AI collaboration and the question of who owns the work
Engineering Performance
A people problem that's actually a systems problem.
Your performance system was designed for people who don't exist.
The annual review assumes a manager who remembers a year accurately, rates without bias, and holds a candid conversation once every twelve months. That manager doesn't exist. We built an entire system around a person we invented.
Look at the data and the system indicts itself. 62% of the variance in performance ratings reflects the rater, not the person being rated. That means most of what your review captures is noise about the manager, not signal about the employee. Then we attach pay, promotion, and someone's sense of their own worth to that noise.
This is the move I make across every topic: stop treating a systems failure as a people failure. Your people aren't bad at performance. Your performance system is bad at measuring it.
The redesign starts by changing what the system is for. Most performance management is built to appraise the past. It should be built to drive the decisions ahead: who to develop, who to stretch, who to promote, what to reward. Stop appraising. Start deciding.
Coaching is the tool that makes it work. A well-designed performance system needs leaders who can set vision and build trust. It needs managers who can actually have the conversations it depends on. That capability isn't charisma. It's a skill set, and it can be taught. That's the argument at the center of my book The Coaching Shift: the best leaders build capability and unlock performance through coaching.
Key ideas
Decision-Driven Performance Management: manage, recognize, promote
Why "bad doors" beat bad people (design the door, not the person)
The true cost of the annual review, and how to calculate it
Coaching as the operating system of modern leadership (*The Coaching Shift*)